Hyperscale power demand, on-site gas generation, battery storage and mining tradeoffs, solid-state batteries, small modular reactors, data centers in space, and agrivoltaics.
Executive Summary
Hyperscale data centers are colliding with power grids faster than new generation can be permitted and built. Iowa is a useful case study because it runs the most wind-heavy grid in the country, yet Microsoft, Google, Meta and Apple facilities there are still straining supply and pushing the state’s largest utility toward new natural gas plants. The pattern repeats nationally: rather than wait years for grid interconnection, some operators — most visibly Elon Musk’s xAI and SpaceX — are installing their own gas turbines on-site, in xAI’s case without air permits, triggering lawsuits and a Justice Department intervention. At the same time, grid-scale batteries are maturing quickly enough to make wind and solar dispatchable on demand, small modular reactors are moving from paper to construction sites in a real but still-unproven industry, and Musk has floated putting data centers in orbit entirely — an idea most independent analysts view as technically serious but economically out of reach for years.
1. Iowa and the Hyperscale Data Center Boom
Iowa has the most wind-powered electricity grid in the United States — wind supplied roughly 63% of the state’s generation in 2024, delivered mainly through MidAmerican Energy’s 27 wind farms. That cheap, clean power, combined with cool climate and flat land, is exactly what drew hyperscale operators to the state: Microsoft’s Project Alluvion in West Des Moines (six buildings, about 200 MW, $5–6 billion invested), Google’s Council Bluffs campus (~$5.5 billion, roughly 95% carbon-free), Meta’s large Altoona campus, and Apple’s facility in Waukee.
The problem is timing, not intent. Data center load is growing faster than new generation and transmission can be built. MidAmerican has said publicly that it expects electricity demand to keep rising “in large part due to data center load growth,” and it has responded with an 800-megawatt Solar Reliability Project (approved, phasing in 2027–2028) and the proposed 518-MW Orient Energy Center — a natural-gas-fired plant. In other words, even in the windiest state in the country, new hyperscale demand is being partly backfilled with gas, because wind and solar alone can’t guarantee power at every hour and new renewable-plus-storage capacity takes years to permit. Some hyperscalers are now bypassing the wait by signing direct power-purchase agreements — Microsoft, for example, contracted 150 MW of dedicated wind power directly — to lock in supply rather than compete for it on the shared grid.
2. On-Site Gas Generation: Musk’s Memphis and Texas Buildout
This is the sharpest example nationally of a hyperscaler skipping the grid queue entirely by generating its own power on-site — and it has become a genuine legal and public-health controversy.
Memphis / South Memphis, Tennessee
xAI’s Colossus supercomputer campus in South Memphis has been running dozens of methane gas turbines — aerial photography identified 35, roughly double what was previously disclosed — without air-pollution permits or public notice. Manufacturer emissions data suggest the turbines have the potential to emit 1,200–2,000 tons per year of smog-forming nitrogen oxides (NOx), which independent estimates say could exceed the output of Memphis International Airport and increase the city’s smog by 30–60%. The turbines also emit formaldehyde, a carcinogen. Boxtown, the majority-Black neighborhood adjacent to the facility, already faces a cancer risk about four times the national average, and residents and advocates argue the turbines are making it worse.
The NAACP, Earthjustice, and the Southern Environmental Law Center have threatened and pursued legal action, arguing xAI is likely the largest single industrial NOx source in Memphis. The U.S. Department of Justice has intervened in the litigation on xAI’s side, citing “national, economic, and energy security” grounds — an unusual move that underscores how much political weight is being put behind AI infrastructure buildout. SpaceX (which now runs the power side of the Colossus 2 site) has committed to phasing out the unpermitted turbines as it builds a permanent, permitted 1.2-gigawatt natural gas plant, but says full turbine removal won’t happen until July 2027 — meaning the unpermitted generation continues for roughly another year.
Texas
There isn’t a single Memphis-style “Colossus Texas” data center; instead, Texas is where Musk’s companies are building out the power-generation supply chain and a new industrial facility, all in-house:
- SpaceX bought Jacksonville-based APR Energy, a fleet of small, trailer-mounted gas turbines and diesel generators that can be installed in days without the standard siting and permitting timeline — a roughly $1 billion move that gives Musk’s companies their own turbine supply for AI and rocket infrastructure.
- SpaceX and Tesla are jointly building “Terafab,” a semiconductor manufacturing plant in Texas (first phase estimated at $16.8 billion, ~3,000 jobs), for which SpaceX will build its own natural-gas power plants and large battery storage systems rather than rely on the ERCOT grid.
- A SpaceX-linked entity is building an 8-mile natural gas pipeline (“Starpipe”) to Starbase in Cameron County to fuel Starship launches more reliably.
- Separately, Texas has become a magnet for merchant gas-fired power built explicitly for data centers — e.g., Pacifico Energy’s GW Ranch in West Texas, permitted for up to 7.7 GW of gas generation to power a private grid for data centers (not a Musk project, but illustrative of the state’s regulatory environment, which approves this kind of large private generation faster than most states).
Read together, the pattern is consistent across both states: Musk’s companies are choosing to own their power generation outright — turbines, pipelines, and now a turbine company — rather than wait on utilities or public grid interconnection queues, and Tennessee/Mississippi and Texas are where that strategy is currently most visible.
3. Grid-Scale Battery Storage: Making Wind and Solar Dispatchable
The core problem with wind and solar is timing — they generate power when the weather cooperates, not necessarily when demand peaks. Batteries solve that by storing excess generation and releasing it on demand, and the technology is scaling quickly. But it isn’t a clean story end to end — the environmental cost mostly shows up upstream, in mining and manufacturing, rather than at the grid site.
Technology and the buildout
Lithium iron phosphate (LFP) chemistry has become the default for utility-scale storage — it’s cheaper and safer than the nickel-based chemistries used in most EVs, at the cost of slightly lower energy density, which matters less for stationary storage. Tesla’s Megapack is the best-known product: Megapack 3, moving to volume production in late 2026, packs 5 MWh per unit (28% more than Megapack 2 in the same footprint), and the new “Megablock” bundles four units with transformers and switchgear into a pre-assembled 20 MWh system designed to cut on-site installation time sharply. Tesla is manufacturing these near Houston, sourcing LFP cells from CATL and increasingly from its own Nevada and Texas production.
The practical effect: pairing storage with renewables turns an intermittent resource into a firm, dispatchable one. Utah’s Green River project combines solar generation with 400 MW of Megapack storage to deliver power on a schedule rather than only when the sun shines.
The pros
Once built, a battery has no emissions and no water use in operation. It replaces gas “peaker” plants that would otherwise run for a few hours a day to cover demand spikes, and it’s what actually lets wind and solar cover demand around the clock rather than only when the weather cooperates. The U.S. is adding a record 24 gigawatts of grid battery storage in 2026, up from a record 15 GW in 2025 — the largest single-year jump in over two decades, evidence this has become a mainstream grid tool rather than a niche one.
The cons: mining, water, and energy
The real environmental cost sits upstream, in the supply chain. Extracting lithium from brine takes roughly 2 million liters of water per ton, causing significant freshwater depletion in the arid regions where most lithium is mined (Chile, Argentina, Australia, and increasingly Nevada). Poorly lined evaporation ponds can also let brine leak into freshwater aquifers, and wastewater from processing facilities can contaminate local water supplies. Manufacturing the battery itself is carbon-intensive — one widely cited estimate found that producing a roughly 1,100-pound battery can emit over 70% more CO2 than manufacturing a conventional car, and about 40% of a battery’s total climate footprint comes from mining and processing the raw minerals alone, before the battery is ever assembled.
Two trends are working against those costs, though. LFP chemistry (the industry standard for grid storage) eliminates cobalt entirely, removing one of the worst-offending minerals from the supply chain. And battery recycling and second-life reuse — taking retired EV batteries and repurposing them for stationary grid storage instead of mining new lithium — is now a real commercial industry: the recycling market alone is projected to roughly triple in size by 2031.
Where it’s being built now, and the biggest projects
Deployment is heavily concentrated in three states, which together account for about 80% of new U.S. capacity in 2026: Texas (53%, ~12.9 GW), California (14%, ~3.4 GW), and Arizona (13%, ~3.2 GW). The largest operating project in the country today is Menifee Power Bank in Riverside, California, at 680 MW. Among projects scheduled to come online in 2026, three of the four largest are in Texas: Lunis Creek BESS in Jackson County (621 MW), Clear Fork Creek Solar and BESS in Wilson (600 MW), and Tehuacana Creek 1 in Navarro County (418 MW), plus Bellefield 2 Solar & Energy Storage in Kern County, California (500 MW).
Iowa’s status
Iowa has proven the concept but hasn’t committed to it at scale. MidAmerican Energy’s only battery project is a small 2019 pilot at a substation in Knoxville — 1 MW / 4 MWh, enough to power roughly 900 homes for four hours, installed to test smoothing wind and solar output. MidAmerican’s big approved generation project, the 800 MW Solar Reliability Project (phasing in 2027–2028), is solar generation only — public filings don’t show a paired battery storage component. So while Texas, California, and Arizona are building gigawatt-scale storage now, Iowa’s storage buildout remains at pilot scale even as its data-center load keeps climbing.
4. Solid-State Batteries: The Next Generation, With Caveats
Solid-state batteries replace the liquid electrolyte in a conventional lithium-ion cell with a solid one, which improves safety (no leakage or flammable liquid), energy density, and lifespan. They are real and in active development, but commercialization is slower than headlines suggest. Toyota is targeting 2027–2028 for commercial deployment, with initial limited production starting in 2026, aiming for roughly 621 miles of range and a 10-minute fast charge in its first-generation pack. QuantumScape is bringing new cell-manufacturing equipment online in February 2026 for its QSE-5 cells. Industry consensus, though, is that large-scale commercialization is unlikely before 2030 — 2026–2030 is considered the critical window for solving cost and manufacturing scale-up, not the finish line, and the technology has so far been developed primarily for electric vehicles rather than grid storage specifically.
Whether solid-state fixes the “cons” of today’s lithium batteries is mixed, not a clean win. Solid-state cells can actually require up to 35% more lithium than current lithium-ion technology, but they use far less graphite and are more likely to adopt cobalt-free chemistries. Their solid electrolyte removes the leakage and contamination risk that liquid electrolytes carry, and their longer lifespan (up to roughly 10 years versus about 2 for current cells) reduces total lifecycle waste and mining demand over time. One life-cycle study found solid-state batteries could cut climate impact by about 39% compared to current lithium batteries — but only if built with the most sustainably sourced materials, which isn’t guaranteed by the technology itself. A genuinely promising parallel development is newer lithium extraction methods, such as recovering it from geothermal wells, which is substantially less destructive than the evaporation-pond brine mining used for most of today’s supply.
5. Small Modular Reactors (SMRs): Who, What, and Where
The SMR sector has moved from concept to early construction, though nothing is yet operating commercially in the U.S. As of mid-2026, at least 66 companies across 15 countries are developing SMR or advanced-reactor designs, roughly 32 of which have real traction — NRC certifications, construction permits, or multi-billion-dollar contracts. The field splits into a few leaders and a much longer tail of earlier-stage developers.
Leading developers
- NuScale Power — the only company with full NRC design certification; announced a partnership with ENTRA1 Energy and the Tennessee Valley Authority for a 6-gigawatt SMR deployment program.
- Oklo — the largest SMR company by market capitalization, part of a Department of Energy fast-track cohort aiming for reactor criticality by July 2026.
- TerraPower (Bill Gates-backed) — broke ground on the first U.S. advanced reactor at Kemmerer, Wyoming.
- Kairos Power — holds the first NRC construction permit issued for an advanced reactor design.
- X-energy, GE Hitachi, and Rolls-Royce SMR — other Tier 1 players with certifications or major contracts in progress.
The Department of Energy separately named 10 companies — Aalo Atomics, Antares Nuclear, Atomic Alchemy, Deep Fission, Last Energy, Oklo, Natura Resources, Radiant Industries, Terrestrial Energy, and Valar Atomics — to a fast-track program targeting demonstration reactors. SMR designs generally range from 30 to 300 MW, small enough to co-locate directly at an industrial site or data center campus rather than requiring a dedicated transmission build-out — which is precisely why hyperscalers are watching the sector closely as a long-term alternative to on-site gas. The caveat: as of early 2026 no Western SMR has begun commercial operation. Russia’s floating Akademik Lomonosov (2020) and China’s HTR-PM (2023) are the only operating examples worldwide, so U.S. timelines remain projections rather than track record.
6. Data Centers in Space: Real Plan, Uncertain Economics
This is genuinely being pursued, not just talked about. SpaceX has filed with the FCC for a constellation of up to roughly 1 million satellites intended to orbit Earth and use solar power to run AI data centers — filed shortly after reports that SpaceX and xAI were discussing a merger ahead of a public offering. A smaller company, Starcloud, has already launched a satellite carrying an Nvidia GPU to test orbital computing. Google has a parallel effort, Project Suncatcher, exploring solar-powered AI satellites, and Axiom Space has discussed sending orbital data-center modules to the ISS by 2027. So the direction is real; the question is scale and timeline.
The case for it
In orbit, solar panels can generate roughly 5 times more power than the same panel on Earth (no atmosphere, no night, no weather), land use is eliminated, and the vacuum of space offers a theoretically clean thermal environment. It also sidesteps the two things terrestrial communities are increasingly pushing back on: grid strain and water consumption for cooling.
The case against it, at least soon
Independent analysts are far more skeptical than the announcements suggest. SemiAnalysis modeled space compute in June 2026 at more than 4 times the cost of terrestrial compute for a comparable GPU cluster ($8.64 vs. $2.37 per GPU-hour). Launch costs are the biggest driver: even a single orbital data center’s cooling radiators — which need roughly 2.15 million square feet of surface area, because radiative cooling in vacuum is over 1,000 times slower than water cooling on Earth — would cost $2–4 billion to launch at today’s prices (~$5,000/kg), or still $200–400 million even at Starship’s target price of ~$500/kg. On top of that, chips in orbit face radiation damage from cosmic rays and space weather that can degrade or kill electronics, and need replacement every 5–6 years, meaning a constant resupply cost most terrestrial data centers don’t have. One widely cited German academic analysis (“Dirty Bits in Low-Earth Orbit,” Saarland University) even argues that once you count launch and re-entry emissions, a solar-powered orbital data center could produce more total emissions than an equivalent land-based one.
Net effect on Earth-side energy and water, if it happens
If orbital data centers ever reach meaningful scale, the clearest terrestrial benefit would be avoided water use — space-based compute needs no cooling water, which is one of the most contentious local issues around new data centers today (including in Iowa, where water use is now drawing the same scrutiny as electricity). The energy picture is murkier: it wouldn’t reduce demand on Earth’s grid so much as redirect it, since building, launching, and replacing orbital hardware is itself extremely energy- and carbon-intensive, and most serious analysts currently expect this to remain a niche, high-cost complement to terrestrial data centers — not a replacement for them — through at least the early 2030s.
7. Agrivoltaics: Combining Solar Farms With Crop Production
Agrivoltaics — also called dual-use solar — means growing crops, grazing livestock, or supporting pollinator habitat underneath and between rows of solar panels rather than treating the land as generation-only. It’s one of the fastest-growing trends in U.S. solar siting, largely because it turns a long-standing objection to solar (“it takes farmland out of production”) into a non-issue, and gives landowners a second revenue stream from the same acre.
Iowa’s own research: Dr. Ajay Nair, Iowa State University
Ajay Nair, chair of Iowa State’s Department of Horticulture, is running a $1.8 million, four-year study funded by the U.S. Department of Energy at the Alliant Energy Solar Farm south of Ames. The project tests how fruits, vegetables, and pollinator plantings grow in the light, temperature, and humidity conditions created underneath and between solar panel rows. First-year results were promising and crop-specific rather than uniformly positive: broccoli came in somewhat smaller than in open-field control plots, but summer squash and peppers performed better under the panels — Nair sees particular promise for reducing sunscald in peppers, a common quality problem in full, unshaded sun. Strawberries and grapes are also part of the trial. This is an active, funded research program, not a one-off pilot.
What’s happening nationally
Solar grazing — running sheep under panel arrays instead of mowing — has become one of the fastest-growing applications, cutting maintenance costs, creating new farm income, and appearing to reduce heat stress in the animals thanks to the shade. Panels also reduce evaporation and help preserve soil moisture, a real advantage in drought-prone areas. Massachusetts has built one of the most developed state-level agrivoltaics ecosystems in the country, backed by land-grant research and extension programs, and Arizona, Oregon, New York, and Vermont are all demonstrating measurable yield benefits for specific crops. Iowa’s entry into this field through Nair’s ISU research is timely: MidAmerican’s 800 MW Solar Reliability Project alone will put a large amount of new solar onto Iowa farmland over the next few years, and agrivoltaics is the clearest tool available for keeping that land in food production rather than taking it out of use entirely.
Sources
Iowa & Hyperscale Data Centers
- Iowa Data Center Power Demand: The Wind-Powered Cluster — gridlas.com
- Energy, water use under close watch as data centers expand in Iowa — Business Record
- How Data Center Growth Is Transforming Iowa’s Cloud and AI Economy — Aureon
- Data Centers in Iowa Resource Hub — Clean Energy Districts
- Data Centers and Electrical Use in Iowa — Iowa’s AI Future
Memphis / xAI Gas Turbines & Litigation
- Elon Musk’s xAI facility is using gas turbines in South Memphis — SELC
- New images reveal xAI datacenter has nearly doubled unpermitted gas turbines — SELC
- xAI explores another massive methane turbine installation at second South Memphis site — SELC
- Inside Memphis’ Battle Against Elon Musk’s xAI Data Center — TIME
- xAI threatened with lawsuit over air pollution, filed on behalf of NAACP — NAACP
- ‘How come I can’t breathe?’ — E&E News / POLITICO
- A billionaire, an AI supercomputer, toxic emissions — Tennessee Lookout
- NAACP Threatens Lawsuit Over xAI’s Unpermitted Gas Turbines in Mississippi — Earthjustice
- SpaceX won’t remove all of xAI’s unpermitted turbines for another year — TechCrunch
- SpaceX to Phase Out Unpermitted Turbines, Building Permanent Gas Plant — AI Insider
- xAI Gets Permits for 15 Natural Gas Generators at Memphis Data Center — TechCrunch
- xAI’s Gas Turbines, the DOJ, and the Data Center Power Permit Maze — Energy.Media
- Colossus (data center) — Wikipedia
Texas Energy Infrastructure
- SpaceX is building a 13-kilometer pipeline in Texas to fuel Starbase — OK Diario
- SpaceX expands into energy infrastructure with Texas pipeline plan — Seeking Alpha
- Elon Musk Discreetly Bought a $1 Billion Gas Turbine Company to Power AI — Motley Fool
- SpaceX to Build Natural Gas Power Plants for Texas Semiconductor Facility — Bloomberg
- Musk abandoned his own ‘solar electric economy’ to burn gas for AI — Electrek
- Space finally faced the people living next to its next Terafab mega-project — Teslarati
- Texas lassoes massive Microsoft datacenter — and 20 years of gas turbine emissions — The Register
Battery Storage
- Tesla Energy Q1 2026 Update: Megapack Deployments Surge to Record Highs — TESMAG
- Elon Musk says Tesla’s energy storage business “very strong” — Energy Storage News
- Tesla just launched the Megablock, a big, easy-to-deploy grid battery — Canary Media
- Megapack — Tesla
- Is Lithium Mining Bad For The Environment? — GreenMatch
- The Environmental Impacts of Lithium and Cobalt Mining — Earth.org
- Environmental Impacts of Lithium-Ion Batteries — Institute for Energy Research
- U.S. battery storage developers plan a record 24 gigawatts in 2026 — GCN
- Top 7 Battery Energy Storage System (BESS) Projects in the USA 2026 — BlackRidge Research
- New US battery capacity in 2026: 24.3 GW of new battery storage — Energy Storage News
- MidAmerican Energy Battery Storage — MidAmericanEnergy.com
- Warren Buffett’s MidAmerican Energy puts in Iowa’s latest big battery project — Energy Storage News
- MidAmerican releases long-term plan to meet projected Iowa demand growth — MidAmericanEnergy.com
- EV Battery Recycling for Investment Recovery: The 2026 Playbook — Invrecovery
- Lithium-Ion Battery Recycling Logistics — CXTMS
Solid-State Batteries
- Solid-State Batteries 2026: How the Technology Is Finally Reaching Commercial Use — to7motor
- Toyota Solid-State Battery Strategy: Patents, Timelines and Reality Check — Future Green Tech
- Solid-state batteries: why mass production won’t happen before 2027 — Christopher Chico
- Solid state batteries can further boost climate benefits of EVs — Transport & Environment
- Environmental Impacts of Solid-State Batteries — Redway Battery
- The Environmental Benefits of Solid-State Batteries — AZoCleantech
Small Modular Reactors
- DOE Names 10 Companies to Fast-Track Small Modular Reactors by 2026 — Energy Tech
- Small Modular Reactor (SMR) Global Project Tracker — World Nuclear Association
- Every Small Modular Reactor Company in 2026: The Complete List — smrintel.com
Data Centers in Space
- Elon Musk’s SpaceX Eyes Solar Data Centers in Space to Power the AI Boom — Carbon Credits
- Elon Musk vows to put data centers in space, run on solar power but experts have doubts — Washington Times
- SpaceX seeks FCC nod for solar-powered satellite data centers for AI — AOL
- Elon Musk confirms SpaceX plans for orbital data centers — Notebookcheck
- Orbital data centers’ feasibility gap is a governance risk — Brookings
- Space-Based Data Centers Could Power AI with Solar Energy — At a Cost — Scientific American
- Four things we’d need to put data centers in space — MIT Technology Review
- Why cooling is the real obstacle to space-based data centres — World Economic Forum
- The Real Cost Of Cooling GPUs In Space Might Shock You — IEEE Spectrum
- Will We Really Put Data Centers in Space? — Forethought
Agrivoltaics
- First Year of Farming at Solar Power Site Shows Promise — Iowa State News Service
- ISU researchers to study growing crops in solar farm’s footprint — Iowa State Research
- ISU researchers say harvesting crops and solar energy don’t have to be at odds — Iowa Public Radio
- Nair announced as next Department of Horticulture chair at Iowa State — CALS
- Department of Energy chief of staff surveys progress at Iowa State solar farm — CALS
- Op-Ed: Agrivoltaics is growing interest across the U.S. — The Clermont Sun
- Agrivoltaics: How combining solar panels and farming delivers big benefits — ASCE
- Harvesting the Sun — Twice: Agrivoltaics and Rural Land-Use — Georgia Tech EPIcenter
